Influence of heart motion on EIT-based stroke volume

estimation

Proença M, Braun F, Rapin M, Solà J, Adler A, Grychtol B,

Bührer M, Krammer P, Bohm SH, Lemay M, Thiran JP; 15th

International Conference on Biomedical Applications of

Electrical Impedance Tomography, Gananoque, Ontario,

Canada, April 24-26, 2014

25 APRIL 2014, 13:30 – 14:45 77

Influence of heart motion on EIT-based stroke volume estimation

Martin Proença1, Fabian Braun

1,5, Michael Rapin

1, Josep Solà

1, Andy Adler

2, Bartłomiej Grychtol

3,

Martin Bührer4, Peter Krammer

5, Stephan H. Bohm

5, Mathieu Lemay

1, Jean-Philippe Thiran

6,7

1Systems Division, Swiss Center for Electronics and Microtechnology (CSEM), Neuchâtel, Switzerland, map@csem.ch 2Systems and Computer Engineering, Carleton University, Ottawa, Canada

3Division of Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany 4Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland

5Swisstom AG, Landquart, Switzerland 6Signal Processing Laboratory (LTS5), Swiss Federal Institute of Technology (EPFL), Lausanne, Switzerland

7Department of Radiology, University Hospital Center (CHUV) and University of Lausanne (UNIL), Lausanne, Switzerland

Abstract: Cardiac electrical impedance tomography (EIT)

signals are affected by myocardial motion. The feasibility

of stroke volume estimation using such signals is thus

questionable. Results based on a dynamic model show that

myocardial motion indeed affects but does not

compromise stroke volume estimation.

1 Introduction

In EIT, cardio-synchronous impedance changes in the

heart region are assumed to reflect variations of blood

volume originating mainly from the ventricles [1]. EIT

appears therefore as an interesting continuous and non-

invasive modality for monitoring total ventricular volume

(TVV), and thus estimating total stroke volume (TSV),

defined as the maximal change in TVV over a full cardiac

cycle. However, there is increasing evidence that other

factors – unrelated to blood volume changes – are

contributing to these variations of cardiac-related

impedance [2]. In that context, simulations we performed

on a finite element 2D extruded dynamic bio-impedance

model showed that EIT signals in the heart region might

be dominated by myocardial motion-induced changes [3].

These findings raised the question whether heart

signals – affected by myocardial motion – remain valid for

estimating changes in TVV and thus TSV. The hypothesis

that the total impedance change in the heart area remains a

true indicator for TVV and TSV is thus investigated here.

2 Methods

To test this hypothesis, we exploited the above 2D

dynamic bio-impedance model – created from segmented

magnetic resonance (MR) data imaged in the heart

horizontal long axis plane – and considered three

scenarios: In Scenario A, we reproduced cardiac blood

volume-related impedance changes by simulating the

filling and emptying of the cardiac cavities. In Scenario B,

myocardial motion-induced changes were reproduced by

simulating the dynamics of the heart muscle. Finally,

Scenario C is the real-case scenario and simulates both

blood volume-related and motion-induced changes [3].

These simulations were performed on our finite

element model over a full cardiac cycle (corresponding to

20 simulated EIT frames) using the open source EIDORS

toolbox, with image reconstruction carried out by the

GREIT approach [4]. For each scenario, the impedance

change ∆ – with respect to end-diastole in the heart area

– was computed for all frames, thus providing an EIT-

based indicator for TVV, according to our hypothesis.

Hereafter referred to as TVVEIT, it was expected to

perform best with Scenario A (no heart motion) and worse

with Scenario B (heart motion only).

The reference TVVREF was obtained by summing the

volumes VLV and VRV of the left and right ventricles. VLV

and VRV were computed via the area-length method [5] –

with the areas (ΣLV and ΣRV) and lengths (LLV and LRV)

coming from the MR data used to create our model:

TVVREF cLV ∙ ΣLV/LLV cRV ∙ ΣRV/LRV ml , (1)

where cLV 8/ [5] and cRV / [6]. LRV was

measured in the vertical long axis plane [6]. The total end-

diastolic volume TEDVREF and the total stroke volume

TSVREF inferred from TVVREF (see Figure 1) were used to

compute TVVEIT by translating ∆ – normalized by its

maximal (systolic) value – into millilitres:

TVVEIT TEDVREF TSVREF ∙ ∆ NORM (ml). (2)

3 Results

The estimation error (mean±SD) between TVVREF and

TVVEIT was of 1.9±13.3, –14.1±19.5 and –10.1±15.7 ml

for Scenario A, B and C, respectively.

Figure 1: Total ventricular volume estimation using simulated

EIT cardiac signals originating from blood volume-related impe-

dance changes (A), motion-induced changes (B), or both (C).

4 Conclusions

In agreement with our expectations, simulations showed

that myocardial motion increased the error on TVVEIT and

thus EIT-based TSV estimation, without however

compromising the approach. When both blood volume

changes and myocardial motion are in action (Scenario C,

real-case scenario) an EIT-based TVV estimation error of

–10.1±15.7 ml was obtained, which is sufficiently low to

be clinically useful in normal subjects [5].

References

[1] Eyüboglu B, et al. IEEE EMBM 8:39-45, 1989

[2] Hellige G, Hahn G. Critical Care 15:430, 2011

[3] Proença M, et al. In BIOSIGNALS. 2014 (in press)

[4] Adler A, et al. Phys Meas 30:S35, 2009

[5] Underwood SR, et al. Br Heart J 60:188-195, 1988

[6] Levine RA, et al. Circulation 69:497-505, 1984

0 2 4 6 8 10 12 14 16 180

50

100

150

200

250

TSVREF

TEDVREF

Frame number (time step = 43 ms)

To

tal

Ven

tric

ula

r V

olu

me (

ml)

TVVREF

TVVEIT

(A)

TVVEIT

(B)

TVVEIT

(C)

S W I S S T O M S C I E N T I F I C L I B R A RY

Co

nte

nt:

Dr.

Ste

ph

an

Bö

hm

; C

on

cep

t &

Desi

gn

: Zw

eiz

eit

Bra

nd

Deve

lop

men

t

Made in Switzerland

electricalimpedancetomography

Real-time tomographic images for organ function monitoring and diagnosis

Contact us!

call: + 41 (0) 81 330 09 72 mail: info@swisstom.comvisit: www.swisstom.com

Swisstom AGSchulstrasse 1, CH-7302 Landquart, Switzerland

Swisstom AG

Swisstom AG, located in Landquart, Switzerland,

develops and manufactures innovative medical devices.

Our new lung function monitor enables life-saving

treatments for patients in intensive care and during

general anesthesia.

Unlike traditional tomography, Swisstom´s bedside

imaging is based on non-radiating principles: Electrical

Impedance Tomography (EIT). To date, no comparable

devices can show such regional organ function

continuously and in real-time at the patient’s bedside.

Swisstom creates its competitive edge by passionate

leadership in non-invasive tomography with the goal

to improve individual lives and therapies.

© Swisstom AG